Tag
Framer's Harness 2 upgrades agent code execution by running JavaScript as it streams, cutting costs by up to 33% and speeding up execution by 20% while reducing tool steps and cached token usage significantly.
TokenCast learns composable cost representations to forecast token consumption in LLM agent execution, reducing prediction error and saving tokens in budget-controlled scenarios.
The article explains the Unified Harness Protocol (UHP) and HarnessRouter, which standardize agent execution across different runtimes like Codex and Claude Code, enabling products to avoid dependency on a single harness.
The article discusses trust boundaries in AI agent systems where an LLM plans an executable DAG of agents, highlighting challenges in validating plans and ensuring safety through human approval and per-agent permissions.
StateM is a runtime system that enhances long-horizon agent execution through durable states and recoverable runbooks, achieving 95.3% accuracy on Terminal-Bench 2.1 and significantly reducing API costs.
This article discusses the differences between Agent execution and general code execution and recommends two sandbox solutions: E2B (based on Firecracker) and OpenSandbox (based on Docker), which are suitable for production-grade and private deployment scenarios, respectively.
This paper introduces OpenClawBench, a large-scale dataset for benchmarking process-side anomalies in real-world AI agent execution trajectories. It reveals that task success can hide process failures, with 9.33% of oracle-passing executions containing anomalies, and provides structured supervision via a novel taxonomy.